Demand Forecasting
Demand forecasting is available on the Pro plan only.
Demand forecasting analyzes your sales history to suggest when and how much to reorder. It helps you maintain optimal stock levels without over-ordering.
How It Works
The forecasting engine uses two key inputs:
- Sales velocity -- how fast a product sells, calculated from historical sales data over a configurable period
- Lead time -- how long it takes to receive new stock from your supplier (configured per supplier or globally)
From these, it calculates:
- Daily sales rate -- average units sold per day
- Days of stock remaining -- current stock divided by daily sales rate
- Suggested reorder quantity -- enough stock to cover the lead time plus a safety buffer
- Reorder urgency -- how soon you need to reorder based on remaining stock vs. lead time
Configuring Forecasting
Analysis Period
Set the number of days of sales history to analyze (e.g., 30, 60, or 90 days). A longer period smooths out anomalies but may miss recent trends. A shorter period is more responsive but can be skewed by unusual weeks.
Lead Time
Set the expected delivery time in days. This can be configured:
- Per supplier -- in the supplier profile within Purchase Orders
- Globally -- as a default for products without a specific supplier lead time
Viewing Suggestions
The forecasting page shows a list of products sorted by reorder urgency:
| Product | Daily Rate | Stock Left | Days Remaining | Lead Time | Suggested Qty |
|---|---|---|---|---|---|
| Widget A | 5.2/day | 15 | 2.9 days | 5 days | 40 |
| Gadget B | 2.1/day | 20 | 9.5 days | 7 days | 25 |
Products where days remaining is less than the lead time are flagged as urgent.
Use forecasting suggestions as a starting point for purchase orders. You can create a PO directly from a forecasting suggestion.
Limitations
- Forecasting is based on historical data and does not account for seasonal trends or planned promotions
- New products with limited sales history will have less accurate predictions
- The model uses a simple moving average -- it does not use machine learning or complex statistical methods